非線性模型預測控制在改良型克勞斯程序之應用

簡易檢索 / 詳目顯示

回結果列表

研究生：	簡浩哲 Chien, Hao-Che
論文名稱：	非線性模型預測控制在改良型克勞斯程序之應用 Application of Nonlinear Model Predictive Control in Modified Claus Process
指導教授：	汪上曉 Wong, David Shan-Hill 姚遠 Yao, Yuan
口試委員:	康嘉麟 Kang, Jia-Lin 劉佳霖 Liu, Jia-Lin
學位類別：	碩士 Master
系所名稱：	工學院 - 化學工程學系 Department of Chemical Engineering
論文出版年：	2024
畢業學年度：	112
語文別：	中文
論文頁數：	95
中文關鍵詞：	非線性程序控制、改良型克勞斯程序、模型預測控制、序列對序列模型
外文關鍵詞：	Nonlinear process control, modified Claus process, Model predictive control, Sequence-to-Sequence
相關次數：	點閱：48 下載：0
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

在鋼鐵工業中，煤炭經過高於1000°C的焦爐進行碳化，生成焦炭。焦爐煤氣（COG）是焦炭製造過程的副產物。COG隨後經過洗滌階段進行脫硫，捕捉到的酸性氣體經過改良型克勞斯程序轉化為元素硫。該製程包括熱反應爐、廢熱鍋爐（WHB）、觸媒反應器（Claus converters）、硫冷凝器和再加熱器。
在熱反應爐中，有三分之一的H2S會與空氣氧化成SO2（H2S + 3/2 O2 → SO2 + H2O），反應溫度通常約在900°C至1400°C之間。依照反應式，增加空氣量可增加消耗 H2S並生成更多SO2，提高H2S的轉化率。然而，增加過多的空氣量，反而因生成過多SO2而消耗過量 H2S，將不利於後段的觸媒反應，造成 H2S 轉化率下降。而熱反應爐出口氣體進入廢熱鍋爐，溫度從1000°C降至300°C。
剩餘的H2S會在較低溫度下（約300°C至200°C）與SO2進行可逆催化反應，並使用Al2O3作為催化劑生成更多的硫(2H2S + SO2 ↔ 1/2 S6 + 2H2O）。每個觸媒反應器後設有冷凝器以凝結元素硫，再透過加熱器將物流重新加熱到反應溫度進入下一個觸媒反應器。而觸媒反應器的操作受到平衡狀態和反應動力學之間的相互作用。當進口溫度低於最佳轉化狀態時，由於SO2 轉換成 S6 的反應為放熱反應，因此提高進料溫度有利於正向反應。然而，反應器溫度高於最佳轉化狀態時,降低進料溫度才有利於觸媒正向反應。
因此在進料溫度低於最佳轉化狀態時，進料溫度與尾氣含硫總濃度有負的過程增益(process gain)；然而，當進料溫度高於最佳轉化狀態時，process gain將變為正值。因此，傳統的比例-積分-微分（PID）控制器無法有效處理觸媒反應器進口溫度的非線性行為，導致操作人員經常將溫度控制器切換為手動模式，進口溫度成為轉換器的潛在變數。此外二次空氣與觸媒反應器進料溫度兩者與尾氣含硫總濃度的時間常數差異巨大。
本研究使用Aspen Plus Dynamics V14對改良型克勞斯程序進行建模，模擬生成了大量的操作數據。使用序列對序列神經網絡(StS)作為代理模型進行非線性模型預測。並且透過AIMPC(模型預測控制)規劃控制策略以最小化出口H2S 及SO2濃度總和，表明深度學習模型在複雜過程控制中具有潛力。

In the steel industry, coal is carbonized to produce coke. Coke oven gas (COG), a by-product consisting of volatile coal substances, is desulfurized through a scrubbing process. The captured sour gas is converted to elemental sulfur using a modified Claus process. This process includes a furnace reactor, a waste heat boiler (WHB), catalytic reactors (Claus converters), a sulfur condenser, and a reheater.
In a thermal reactor, one-third of H2S is oxidized to SO2 with air (H2S + 3/2 O2 → SO2 + H2O). Increasing the air amount can increase H₂S consumption and produce more SO₂, thus increasing the H₂S conversion rate. However, adding too much air will produce excess SO₂ , which consumes too much H₂S. This harms the subsequent catalytic reaction and lowers the H₂S conversion rate. The outlet gas enters the WHB, reducing the temperature from 1000°C to 300°C.
In the catalytic reactor, the remaining H2S reacts with SO2 to produce more sulfur in a reversible reaction (2H2S + SO2 ↔ 1/2 S6 + 2H2O). A condenser is installed after each catalyst reactor to condense the elemental sulfur, and reheaters are used to reheat.
The operation of catalytic reactors is constrained by equilibrium states and reaction kinetics. When the inlet temperature is lower than optimal, increasing the feed temperature benefits the forward reaction, as the conversion of SO₂ to S₆ is exothermic. However, when the reactor temperature is higher than optimal, lowering the feed temperature is advantageous.
Traditional Proportional-Integral-Derivative (PID) controllers cannot effectively manage the nonlinear behavior of the inlet temperature, leading operators to switch to manual mode. Additionally, the secondary air (Air2) flow has a nonlinear impact on the total sulfur concentration in the tail gas, and there is a significant time delay in manipulating airflow to H2S conversion.
In this study, a modified Claus process was modeled using Aspen Plus Dynamics, generating extensive operational data to reflect typical steel plant production conditions. A sequence-to-sequence (StS) neural network was employed as a surrogate model for nonlinear model predictive control (NLMPC). The results demonstrate that this approach effectively minimizes H2S and SO2 concentrations, highlighting the significant potential of deep learning models in complex process control.

中文摘要    i
Abstract    ii
總目錄    iii
圖目錄    v
表目錄    viii
第一章 緒論    1
1 研究背景    1
2 文獻回顧    4
3 動機、目標及章節安排    5
第二章 穩態Aspen plus模型建立    6
1 熱力學與動力學模型及文獻驗證    6
1.1 反應動力學設定    6
1.2 文獻驗證    12
2 中鋼Modified Claus穩態製程模擬    15
2.1 流程說明    15
2.2 ASPEN 模擬    16
2.3 反應器設定    18
2.3.1 Burner及Furnace區    18
2.3.2 觸媒反應區    19
2.4 熱回收設定    20
2.4.1 E-WHB    20
2.4.2 B8    21
2.5 進料條件    22
2.6 優化操作    27
第三章 建立Aspen Dynamics動態模型    30
1 前言    30
2 動態模擬設定    30
3 控制迴圈架構    34
3.1 多變數控制    35
3.2 PID參數設定    40
4 動態控制結果    42
4.1 進料酸氣+5%    42
4.2 進料酸氣-5%    44
4.3 系統動態特徵    46
4.4 傳統控制與工廠現行狀況比較    52
第四章 序列對序列神經網路建立    54
1 理論    54
1.1 人工神經網路    54
1.2 激活函數    55
1.3 遞歸神經網路（Recurrent Neural Network, RNN）    56
1.4 門控循環單元（Gated Recurrent Unit, GRU）    58
1.5 序列對序列神經網路架構(Sequence to Sequence)    59
1.6 Claus製程之StS模型    60
2 數據產生    62
2.1 進料組成    62
2.2 資料前處理    65
2.3 資料畫分    65
2.4 模型評判標準    66
3 模型架構選擇    67
3.1 模型架構選擇    67
3.2 Unit cells單元數量(N)    68
3.3 編碼器及解碼器長度(W，H)    71
3.4 模型架構    72
第五章 模型預測控制    81
1 模型預測控制    81
2 控制策略    84
2.1 729種控制策略    85
2.2 9種控制策略    89
第六章 結論    92
第七章 參考文獻    93




 

                                

1. Liu, J., Tsai, B. Y., & Chen, D. S. (2023). Deep reinforcement learning based controller with dynamic feature extraction for an industrial claus process. Journal of the Taiwan Institute of Chemical Engineers, 146, 104779.
2. Nabikandi, N. J., & Fatemi, S. (2015). Kinetic modelling of a commercial sulfur recovery unit based on Claus straight through process: Comparison with equilibrium model. Journal of Industrial and Engineering Chemistry, 30, 50-63.
3. Clark, P. D., Dowling, N. I., & Huang, M. (2001). Conversion of CS2 and COS over alumina and titania under Claus process conditions: reaction with H2O and SO2. Applied Catalysis B: Environmental, 31(2), 107-112.
4. Hawboldt, K. A. (1998). Kinetic modelling of key reactions in the modified Claus plant front end furnace (Doctoral dissertation, Ph. D. Thesis, Department of Chemical and Petroleum Engineering).
5. Contributors, E. (2012). Gas Processors Suppliers Association (GPSA). FPS VERSION edn., Gas Processors Suppliers Association (GPSA), Tulsa, Oklahoma, 1-34.
6. Selim, H., Ibrahim, S., Al Shoaibi, A., & Gupta, A. K. (2013). Effect of oxygen enrichment on acid gas combustion in hydrogen/air flames under claus conditions. Applied energy, 109, 119-124.)
7. Ibrahim, S., Al Shoaibi, A., & Gupta, A. K. (2015). Xylene addition effects to H2S combustion under Claus condition. Fuel, 150, 1-7.
8. An, S., & Jung, J. C. (2020). Kinetic modeling of thermal reactor in Claus process using CHEMKIN-PRO software. Case Studies in Thermal Engineering, 21, 100694.
9. Monnery, W. D., Svrcek, W. Y., & Behie, L. A. (1993). Modelling the modified claus process reaction furnace and the implications on plant design and recovery. The Canadian Journal of Chemical Engineering, 71(5), 711-724.
10. Manenti, F., Papasidero, D., Frassoldati, A., Bozzano, G., Pierucci, S., & Ranzi, E. (2013). Multi-scale modeling of Claus thermal furnace and waste heat boiler using detailed kinetics. Computers & Chemical Engineering, 59, 219-225.
11. Manenti, F., Papasidero, D., Bozzano, G., & Ranzi, E. (2014). Model-based optimization of sulfur recovery units. Computers & chemical engineering, 66, 244-251.
12. Zarei, S., Ganji, H., Sadi, M., & Rashidzadeh, M. (2016). Kinetic modeling and optimization of Claus reaction furnace. Journal of Natural Gas Science and Engineering, 31, 747-757.
13. Jones, D. D. (2011). Steady State and Dynamic Modeling of the Modified Claus Process as Part of an IGCC Power Plant. West Virginia University.
14. Monnery, W. D., Hawboldt, K. A., Pollock, A., & Svrcek, W. Y. (2000). New experimental data and kinetic rate expression for the Claus reaction. Chemical Engineering Science, 55(21), 5141-5148.
15. Karan, K., & Behie, L. A. (2004). CS2 formation in the claus reaction furnace: a kinetic study of methane− sulfur and methane− hydrogen sulfide reactions. Industrial & engineering chemistry research, 43(13), 3304-3313.
16. Tong, S., Dalla Lana, I. G., & Chuang, K. T. (1997). Effect of catalyst shape on the hydrolysis of COS and CS2 in a simulated Claus converter. Industrial & engineering chemistry research, 36(10), 4087-4093.
17. Sassi, M., Gupta, A. K. (2008). Sulfur recovery from acid gas using the Claus process and high temperature air combustion (HiTAC) technology. American Journal of Environmental Sciences, 4(5), 502-511.
18. Willis, M. J. (1999). Proportional-integral-derivative control. Dept. of Chemical and Process Engineering University of Newcastle, 6.
19. J.Han, C.Moraga, (1995) The influence of the sigmoid function parameters on the speed of backpropagation learning, In International Workshop on Artificial Neural Networks, 195-201.
20. V.Nair, G.E.Hinton, (2010), Rectified linear units improve restricted boltzmann machines, In ICML, 807-814.
21. M.Hermans, B.Schrauwen, (2013) Training and analysing deep recurrent neural networks, In Advances in neural information processing systems, 190-198.
22. Cho, K., Van Merriënboer, B., Bahdanau, D., & Bengio, Y. (2014). On the properties of neural machine translation: Encoder-decoder approaches. arXiv preprint arXiv:1409.1259.
23. Schwaller, P., Gaudin, T., Lanyi, D., Bekas, C., & Laino, T. (2018). “Found in Translation”: predicting outcomes of complex organic chemistry reactions using neural sequence-to-sequence models. Chemical science, 9(28), 6091-6098.
24. Sutskever, I., Vinyals, O., & Le, Q. V. (2014). Sequence to sequence learning with neural networks. Advances in neural information processing systems, 27.
25. Babcock, B., Datar, M., & Motwani, R. (2002, January). Sampling from a moving window over streaming data. In Proceedings of the thirteenth annual ACM-SIAM symposium on Discrete algorithms (pp. 633-634).
26. Garcia, C. E., Prett, D. M., & Morari, M. (1989). Model predictive control: Theory and practice—A survey. Automatica, 25(3), 335-348.
27. Holkar, K. S., & Waghmare, L. M. (2010). An overview of model predictive control. International Journal of control and automation, 3(4), 47-63.
28. Wu, Z., Rincon, D., Luo, J., & Christofides, P. D. (2021). Machine learning modeling and predictive control of nonlinear processes using noisy data. AIChE Journal, 67(4), e17164.
29. Bindlish, R. (2015). Nonlinear model predictive control of an industrial polymerization process. Computers & Chemical Engineering, 73, 43-48.
30. Rho, H. J., Huh, Y. J., & Rhee, H. K. (1998). Application of adaptive model-predictive control to a batch MMA polymerization reactor. Chemical Engineering Science, 53(21), 3729-3739.
31. Lucia, S., Finkler, T., & Engell, S. (2013). Multi-stage nonlinear model predictive control applied to a semi-batch polymerization reactor under uncertainty. Journal of process control, 23(9), 1306-1319.
32. Abellan-Nebot, J. V., & Romero Subirón, F. (2010). A review of machining monitoring systems based on artificial intelligence process models. The International Journal of Advanced Manufacturing Technology, 47, 237-257.
33. Kalogirou, S. A. (2003). Artificial intelligence for the modeling and control of combustion processes: a review. Progress in energy and combustion science, 29(6), 515-566.

簡易檢索 / 詳目顯示

相關論文